Agentmemory
The official website positions Agentmemory as a local persistent memory product for AI coding agents, claiming to capture sessions and recall them in subsequent sessions.
Who it is for: Developers using coding agents that support hooks, MCP, or REST API
Core capabilities
- Automatic Session Capture: The product provides 12 automatically captured hooks, sending session start, prompt, tool call, and session end events into the memory pipeline.
- Hybrid Retrieval and Version Governance: The official site states that the main retrieval path combines BM25, vector, and knowledge graph scores and re-ranks them on-device; replaced versions do not appear in current retrieval results, but historical records are still queryable.
- Local Execution: The official website and documentation state that the product runs locally and claims no external database is needed.
- Session Integration: Official documentation states that after configuring provider key, compression, summarization, integration, and graph extraction are enabled; the official site states that when a session ends, raw observations can be compressed, duplicates merged, outdated records decayed, and audit lines recorded.
- Session Import and Extraction: The official site states that the product can import Claude Code JSONL sessions, rebuild indexes, and generate session crystals and lessons from them.
Pricing
Only the free signal can be confirmed; package details, permanent free boundaries, enterprise support, or pay-as-you-go pricing cannot be confirmed.
Updated 2026-08-18